If your OpenAI spend is over $1,000 per month, DM me. 👀

Text2Video-Zero: Text-to-Image Diffusion Models are Zero-Shot Video Generators abs: arxiv.org/abs/2303.13439 github: github.com/Picsart-AI-Res…

AI progress has skyrocketed with the release of #ChatGPT 🚀 As AI systems get complex, the ability to effectively debug and monitor them becomes crucial. Check out my latest article on how to use AimStack to monitor AI systems built with bit.ly/3Kk7Xwl

🚀 Building Multi task AI agent with and using AimStack to trace and visualize the executions Let’s play.. We’ll prompt the agent to discover who Leonardo DiCaprio’s girlfriend is and calculate her current age raised to the power of 0.43🧵 1/7

🇬🇧 In London? Want to learn more about machine translation quality prediction? Meet the ModelFront team in London this week for XTM Live, the The Machine Translate Foundation meetup and LocLunch

Our next machine translation meetup will be in London! Join us tomorrow April 18th, 4-6 pm, to enjoy drinks, pizza and fun conversations. Sign up at machinetranslate.org/meetup

.ModelFront's Adam Bittlingmayer on translation quality prediction 🎙 Catch up on slator.ch/Pod162 📺 Or head over to Apple Podcasts 👉bit.ly/slatorpod and Spotify 👉bit.ly/slator-pod. Adam Bittlingmayer #t9n #xl8 #AI #tech #LLM #LLMs

Adam Bittlingmayer, CEO of ModelFront 🎙️ on how LLMs are changing the language industry landscape. Adam Bittlingmayer #LLM #ChatGPT #translation #xl8 #t9n youtu.be/WB23OF27Bw0

People severely underestimate the complexity of building LLM's. Anyone that has trained good LLM's will tell you it's non-trivial. There are likely a couple hundred people in the world that could train models competitive with ≥GPT-3.5.

This is a thread about what was right and what was wrong in the recent Senate hearings on AI harms featuring OpenAI CEO Sam Altman, as well as IBM's Christina Montgomery and Gary Marcus (one of the researchers in this field I respect most).

Susan & I found MMLU performance jump 6-10 points in the 40s by formatting multiple choice as (A) not A in MMLU (for internal model). All evaluation of LLM's are broken. Evaluating a task requires marginalizing across all prompts that describe the task, not point estimate of one.

Another example of how ChatGPT has caused more tasks to follow the shift that already happened in translation since 2016.

Narrative: all AI is in SF! Cerebral valley! Counter-narrative: * #1 open source LLM: Falcon, built in Abu Dhabi, UAE * It's trending on HuggingFace, a NYC-based company (/France) * Big AI funding news of the week: Synthesia (London) and Mistral (Paris) AI = truly global.

Founders! If you're building on Stability AI models, please DM me – I'd love to share what we've been working on 💫

New!! "ServiceTitan: The First Great SaaS IPO of 2024 (Potentially) at $500,000,000+ ARR" saastr.com/servicetitan-t… saastr.com/servicetitan-t…

How is AI accelerating translation for the leading global luxury marketplace? 📈 At XTM Live, Alex Katsambas will share the story of how Farfetch rolled out AI to accelerate human-quality translation by ██%. April 11th in NY 🗽

[1/7] 🚀 Introducing the Language Model Transparency Tool - an open-source interactive toolkit for analyzing Transformer-based language models. We can't wait to see how the community will use this tool! github.com/facebookresear…

We’re sharing a **practical guide** to machine translation **quality prediction** with you, based on what we’ve learned making “quality estimation” research work in the real world.